Do longer duration batteries stack up?

TL;DR

Inspired by RWE’s recently announced long duration battery at their Limondale solar farm in NSW, Australia, we modelled a two hour and an eight hour battery to see which makes the better investment. 

What did we learn? The longer duration asset makes sense, but only in the scenario where it’s coupled to a heavily curtailed renewable generation asset. Outside of that particular setup, or in the absence of new value streams that incentivise longer duration batteries, the additional capex investment doesn’t stack up.

Batteries are typically short duration

Since their initial deployments almost 10 years ago, large grid-scale batteries have generally focused on high-power, short-duration applications. For instance, the first Hornsdale battery in South Australia had a capacity of 100MW/125MWh, yielding a duration of just 1.25 hours. Subsequent storage projects, both in Australia and abroad, typically sit within the one to two-hour duration range.

This trend primarily relates to the roles these assets were intended to play in the market. The primary focus has been on grid support services, especially post-fault or contingency frequency control. Frequency control services are more "power-limited" than "energy-limited," making the battery's capacity in MW significantly more important than its energy storage in MWh.

But that pattern looks like it could be starting to change and the recently announced eight hour duration project from RWE at their Limondale solar farm surprised the industry by beating pumped hydro in a recent NSW Government tender.

There are a few things driving the shift:

1. Traditional value streams are drying up: post-fault frequency control has quickly become saturated in international markets (and is broadly predicted to in the NEM as well) meaning that revenue streams for assets designed specifically for this purpose are drying up. That’s forcing asset owners to look elsewhere to turn a dollar and arbitraging the energy markets or providing regulation frequency control are two obvious places to play.

2. Shoring up existing renewable assets: owners of existing renewable energy projects such as solar farms and wind farms are increasingly suffering from either commercial or physical curtailment that’s trimming their revenues. Commercial curtailment might be caused by negative wholesale market prices; physical curtailment by congestion on the transmission network preventing them moving their energy to market.

3. Capacity market derating: for markets that include a capacity mechanism, for example Great Britain or Western Australia’s WEM, capacity payments are typically limited for shorter duration batteries on the basis that a one hour battery isn’t as useful in terms of providing capacity to the market as longer duration assets

4. Market operators and grid balancing: as the shift to renewable energy continues apace market operators are increasingly focussed on intra and interday balancing of renewable generation to demand and are starting to develop new market mechanisms to incentivise investment in assets that can alleviate the inevitable mismatch between periods of high solar and wind generation and periods of relative scarcity.

So, if battery durations are increasing from an hour or two out to eight hours in the case of Limondale, will these longer duration assets make more money than their temporally challenged counterparts, and how might you work that out?

Context

Limondale is an existing 249MWac / 349MWdc solar farm located in south western NSW in Australia’s National Electricity Market (NEM). It’s exposed to the NEM NSW wholesale price and is located in a pretty constrained part of the transmission network.

A combination of commercial and physical constraints mean that in 2022 it was curtailed for one reason or another to the tune of about 20% of its output**. To put that another way, for every 10 MWh of energy the plant generated, 2MWh were spilled and so couldn’t be sold.

In May 2023 RWE, the owners of the solar farm, announced they were building an eight hour (50MW/400MWh) battery at the site.

So that’s the jump off point for the simulation.

Modelling Setup

In this case we’ve chosen to run a retrospective simulation to see what kind of value the eight hour battery might have delivered in 2022 based on the actual conditions experienced over the year, including weather, commercial and physical constraints** and wholesale market prices. 

It's worth remembering that 2022 was a wild year in the NEM with record high prices and including a market suspension. It was also a strong La Nina year with lower levels of sunshine than normal for the east coast. 

But whilst 2022 is (hopefully) not representative of future years to come, the modelling still reveals the relative performance of the batteries, and that is something that is useful in terms of understanding the rationale for investment in long duration assets. 

So we’ve then compared the revenue from the longer duration battery to revenues for a two hour (50MW/100MWh) battery to see how much better the long duration asset does. This should hopefully provide some insights into why RWE went with such a long-duration battery.

In terms of value streams, we’re assuming both batteries are only chasing wholesale energy revenue, not frequency control services. The rationale there is that longer duration batteries are all about moving more energy. As mentioned up top, frequency markets, certainly contingency frequency markets, are mostly about power, aka capacity, and not energy. There's  also no consideration of potential value associated with the NSW Government's long duration storage auction.

Practically speaking, that means the focus is on arbitraging the wholesale price as well as helping to capture the curtailed solar output from the solar farm that would otherwise be lost.

Modelling Results

Starting with the financials, total revenue for all three scenarios broken down by asset value stream below shows the eight hour battery collects $49MM compared to the two hour BESS collecting $13M, so for 4x the storage capacity we’ve got 3.7x the revenue. 

The significant majority of revenue (dark blue) is earned from selling that curtailed solar i.e. the battery charges off “free” solar that would otherwise be unable to get to market and then discharges when market prices and network access are favourable. The rest of the battery revenue (light blue) is earned through wholesale arbitrage, so charging from the grid when prices are low and discharging when they’re high.

Looking at the monthly breakdown of battery revenue it’s easy to see just how volatile 2022 was with the majority of revenue coming from just May, June and July; peak bedlam in the NEM. 

It’s noteworthy that through the second half of the year, with volatility starting to settle, revenue from curtailed solar stood up much better than that from wholesale arbitrage. Access to that curtailed solar is of course a value stream only available to batteries that are co-located with generation.

Moving to the physical energy flows, in the baseline scenario the solar farm generates about 674 GWh over the course of the year, of which 579GWh is exported to the grid and 95 GWh (15% of gross generation) is curtailed.

In the battery scenarios the two hour battery is able to capture 20GWh (21%) of that curtailed energy and the eight hour battery captures 61GWh (64%). So a 4x increase in storage capacity yields a 3x increase in curtailed solar energy that can instead be stored and sold.

Drilling deeper into the 5-minute resolution interval data we can see the performance of the site for each scenario for an example week February 2022

Solar generation is shown in yellow, energy flows through the grid connection in blue and battery state-of-charge in purple. Notice on Thursday Feb 18th and Friday Feb 19th there was considerable curtailment of solar energy in the baseline scenario (the blue grid export is not symmetrical to the yellow solar generation), and that subsequently the batteries were able to charge off this energy and discharge it into the market after dark. Also notice that the two hour battery is only able to capture a modest amount of curtailed solar output before it’s full.

‍

 So do we have a winner?

Using CSIROs most recent Gencosts report*** we can guestimate the capex at ~$65MM for the two hour battery and ~$195MM for the eight hour version, so that’s 3x the cost for 4x the energy storage and 3.7x the one-year revenue ($49MM vs $13MM)

Looked at through that very reductive lens, the eight hour battery does seem like a goer, and that’s before we consider other potential revenue streams that might particularly favour longer duration assets. 

Looking ahead to the next few years it’s hard to imagine another year like 2022 and certainly all the market forecasts see prices falling steadily over the next decade, albeit with continued periods of significant volatility. 

It’s easy enough to imagine owners of existing constrained generation assets considering long duration storage as the business case is less reliant on high levels of volatility (and equally high trading prowess). For standalone storage assets you might think some new value streams need to crystallise a little more before it makes sense for them.

Small print
** curtailment estimate based on AEMO dispatch data, specifically the difference between MW available and MW cleared.

*** https://publications.csiro.au/rpr/download?pid=csiro:EP2022-5511&dsid=DS1

‍

‍